Re: [GIT PULL] Kbuild fixes for v5.19-rc4

[Masahiro Yamada]

Hi Linus,


On Mon, Jun 27, 2022 at 3:20 AM Linus Torvalds
<torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
>
> On Sat, Jun 25, 2022 at 3:16 PM Masahiro Yamada <masahiroy@xxxxxxxxxx> wrote:
> >
> > Please pull a couple of Kbuild fixes.
>
> Masahiro, I already complained about this once:
>
>     Section mismatch in reference from the variable .. declared __init
>
> and now you re-introduced this sneakily by just putting it in another place.
>
> I told you last time that you need to *FIX* these things before adding
> the warning, instead of adding a warning and then waiting for others
> to fix them.

Sorry for not explaining this in my email.

I just thought I had fixed all the warnings,
at least the ones previously reported in linux-next:
https://lore.kernel.org/all/20220602092809.04008a86@xxxxxxxxxxxxxxxx/




> Yes, it's fixing an earlier mistake, but that earlier mistake is over
> a decade old by now.
>
> Now, the only warning that seems to show up for me is about
> tick_nohz_full_setup. Which is only called from housekeeping_setup()
> in kernel/sched/isolation.c if I grepped correctly.


Sorry, I did not notice this.
It did not show up in allmodconfig test.
tick_nohz_full_setup() is only compiled when CONFIG_NO_HZ_FULL=y.

I rely on the 0day bot for randconfig tests, but I did not get the report.


Sorry about that.


>
> And that seems only ever built-in as far as I can tell, so the whole
> export seems pointless, and the fix is presumably to remove the
> export_SYMBOL_GPL entirely.
>
> So it seems fixable.
>
> But I'm upset about this sneaking in when we _talked_ about this issue earlier.
>
> Paul - that export was added by commit ae9e557b5be2 ("time: Export
> tick start/stop functions for rcutorture"), and it seems to have been
> bogus even at that time.
>
>               Linus


I can send a patch.


-- 
Best Regards
Masahiro Yamada